LangChain Releases Open-Source Agent 'Deep Life Sci' for Life Sciences Research

LangChain has released Deep Life Sci, an open-source agent template for clinical and laboratory scientists. This solution addresses the issue of Eroom's law, where new drug development costs double every nine years, and reflects the 'Own your intelligence' philosophy, allowing enterprises to control their own data and context.

Deep Life Sci is built on the Deep Agents harness and can directly reference over 600,000 studies from ClinicalTrials.gov, as well as 29 million abstracts and 12 million full texts from PubMed. Each agent can process files in various formats such as PDF, FASTA, and SMILES within a LangSmith sandbox, execute arbitrary analysis code, and utilize hundreds of sub-agents to review large documents in parallel.

Key supported workflows include:

• RNA-seq/Proteomics analysis: Perform enrichment on uploaded data and identify differentially expressed genes
• Clinical/HEOR comparison: Extract clinical trial data for standard treatments to generate forest-plot style comparison tables
• Clinical documentation: Assist in auditing and revising thousands of pages of regulatory documents such as informed consent forms and protocols

This template applies the Agent Development Lifecycle (ADLC) to support development, testing, deployment, and monitoring loops. All execution processes are logged end-to-end in LangSmith for debugging and audit trails, and the built-in eval set allows verification of performance changes and silent regressions when swapping models or modifying prompts. Enterprises can customize the open-source code to meet their own requirements, such as integrating internal data, swapping models, and applying security guardrails.